static int save_dataset_comments (const dataset *dset)
{
int i, j;
cst_voice *v;
cst_wave *w, *fullw = NULL;
flite_init();
v = register_cmu_us_kal();
j = 0;
for (i=0; i<N_COMMENTS; i++) {
if (dset->comments[i] != NULL) {
if (j == 0) {
fullw = flite_text_to_wave(dset->comments[i], v);
} else {
w = flite_text_to_wave(dset->comments[i], v);
concat_wave(fullw, w);
delete_wave(w);
}
j++;
}
}
cst_wave_save_riff(fullw, "gretl_flite.wav");
delete_wave(fullw);
return 0;
}
float flite_process_output(cst_utterance *u, const char *outtype,
int append)
{
/* Play or save (append) output to output file */
cst_wave *w;
float dur;
if (!u) return 0.0;
w = utt_wave(u);
dur = (float)w->num_samples/(float)w->sample_rate;
if (cst_streq(outtype,"play"))
play_wave(w);
else if (cst_streq(outtype,"stream"))
{
/* It's already been played so do nothing */
}
else if (!cst_streq(outtype,"none"))
{
if (append)
cst_wave_append_riff(w,outtype);
else
cst_wave_save_riff(w,outtype);
}
return dur;
}
int main(int argc, char **argv)
{
cst_wave *nw, *all;
cst_val *files;
const cst_val *w;
cst_val *wavelist;
cst_features *args;
int i,j;
float ntime;
int stime;
const char *nwfile;
args = new_features();
files =
cst_args(argv,argc,
"usage: combine_waves OPTIONS\n"
"Combine waves into single waveform\n"
"-o <string> Output waveform\n"
"-f <int> Input sample rate (for raw input)\n"
"-itype <string> Input type, raw or headered\n"
"-wavelist <string> File containing times and wave filenames\n",
args);
wavelist = get_wavelist(get_param_string(args,"-wavelist","-"));
if (wavelist == 0)
return -1;
all = new_wave();
for (w = wavelist; w; w = val_cdr(w))
{
ntime = decode_time(val_string(val_car(w)));
nwfile = val_string(val_car(val_cdr(w)));
nw = new_wave();
if (cst_wave_load_riff(nw,nwfile) != CST_OK_FORMAT)
{
fprintf(stderr,
"combine_waves: can't read file or wrong format \"%s\"\n",
nwfile);
continue;
}
stime = ntime * nw->sample_rate;
cst_wave_resize(all,stime+nw->num_samples,1);
for (i=0,j=stime; i<nw->num_samples; i++,j++)
{
/* this will cause overflows */
all->samples[j] += nw->samples[i];
}
delete_wave(nw);
}
cst_wave_save_riff(all,get_param_string(args,"-o","-"));
return 0;
}
int flowm_save_wave(TCHAR *filename)
{
/* Save the Last synthesized waveform file to filename */
char *sfilename;
int rc;
if (!previous_wave)
return -1;
sfilename = cst_wstr2cstr(filename);
rc = cst_wave_save_riff(previous_wave,sfilename);
cst_free(sfilename);
return rc;
}
float flite_ssml_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_tokenstream *ts;
int fp;
cst_wave *w;
float d;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for ssml reading\n",
filename);
return 1;
}
fp = get_param_int(voice->features,"file_start_position",0);
if (fp > 0)
ts_set_stream_pos(ts,fp);
/* If its a file to write to, create and save an empty wave file */
/* as we are going to incrementally append to it */
if (!cst_streq(outtype,"play") &&
!cst_streq(outtype,"none") &&
!cst_streq(outtype,"stream"))
{
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
d = flite_ssml_to_speech_ts(ts,voice,outtype);
ts_close(ts);
return d;
}
int main(int argc, char **argv)
{
cst_wave *in, *out;
cst_val *files;
cst_features *args;
int i,j;
int w, a, t;
args = new_features();
files =
cst_args(argv,argc,
"usage: dcoffset_wave OPTIONS\n"
"Subtract window average from waveform\n"
"-i <string> Input waveform\n"
"-o <string> Output waveform\n"
"-w <int> Window size (in samples)\n",
args);
w = flite_get_param_int(args,"-w",20);
in = new_wave();
cst_wave_load_riff(in,flite_get_param_string(args,"-i","-"));
out = copy_wave(in);
for (i=0; i<=out->num_samples; i++)
{
for (t=a=0,j=i-w/2; j < i+w/2; j++)
{
if ((j > 0) && (j < out->num_samples))
{
t += 1;
a+=in->samples[j];
}
}
/* printf("%d %d %d %d %d\n",i,out->samples[i],a/t,t,out->samples[i]-a/t); */
out->samples[i] -= a/t;
}
cst_wave_save_riff(out,flite_get_param_string(args,"-o","-"));
return 0;
}
int main(int argc, char **argv)
{
cst_track *lpc;
cst_wave *sig, *sig2;
cst_sts *sts;
if (argc != 6)
{
fprintf(stderr,"usage: find_sts lpc_min lpc_range LPC WAVEFILE STS\n");
return 1;
}
lpc_min = atof(argv[1]);
lpc_range = atof(argv[2]);
lpc = new_track();
cst_track_load_est(lpc,argv[3]);
sig = new_wave();
if (cst_wave_load_riff(sig,argv[4]) == CST_WRONG_FORMAT)
{
fprintf(stderr,
"cannot load waveform, format unrecognized, from \"%s\"\n",
argv[4]);
exit(-1);
}
sts = find_sts(sig,lpc);
/* See if it worked */
sig2 = reconstruct_wave(sig,sts,lpc);
compare_waves(sig,sig2);
cst_wave_save_riff(sig2,"sig2.wav");
save_sts(sts,lpc,sig,argv[5]);
return 0;
}
float flite_phones_to_speech(const char *text,
cst_voice *voice,
const char *outtype)
{
cst_utterance *u;
cst_wave *w;
float durs;
u = flite_synth_phones(text,voice);
if (u == NULL)
return -1;
w = utt_wave(u);
durs = (float)w->num_samples/(float)w->sample_rate;
if (cst_streq(outtype,"play"))
play_wave(w);
else if (!cst_streq(outtype,"none"))
cst_wave_save_riff(w,outtype);
delete_utterance(u);
return durs;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float d, durs = 0;
int num_tokens;
cst_breakfunc breakfunc = default_utt_break;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
/* If its a file to write to delete it as we're going to */
/* incrementally append to it */
if (!cst_streq(outtype,"play") && !cst_streq(outtype,"none"))
{
cst_wave *w;
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt */
d = flite_tokens_to_speech(utt,voice,outtype);
utt = NULL;
if (d < 0)
goto out;
durs += d;
if (ts_eof(ts))
goto out;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
item_set_int(t,"file_pos",ts->file_pos);
item_set_int(t,"line_number",ts->line_number);
}
out:
delete_utterance(utt);
ts_close(ts);
return durs;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float durs = 0;
int num_tokens;
cst_wave *w;
cst_breakfunc breakfunc = default_utt_break;
cst_uttfunc utt_user_callback = 0;
int fp;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
fp = get_param_int(voice->features,"file_start_position",0);
if (fp > 0)
ts_set_stream_pos(ts,fp);
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
if (feat_present(voice->features,"utt_user_callback"))
utt_user_callback = val_uttfunc(feat_val(voice->features,"utt_user_callback"));
/* If its a file to write to, create and save an empty wave file */
/* as we are going to incrementally append to it */
if (!cst_streq(outtype,"play") &&
!cst_streq(outtype,"none") &&
!cst_streq(outtype,"stream"))
{
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((cst_strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt, so synthesize it */
if (utt_user_callback)
utt = (utt_user_callback)(utt);
if (utt)
{
utt = flite_do_synth(utt,voice,utt_synth_tokens);
durs += flite_process_output(utt,outtype,TRUE);
delete_utterance(utt); utt = NULL;
}
else
break;
if (ts_eof(ts)) break;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
/* Mark it at the beginning of the token */
item_set_int(t,"file_pos",
ts->file_pos-(1+ /* as we are already on the next char */
cst_strlen(token)+
cst_strlen(ts->prepunctuation)+
cst_strlen(ts->postpunctuation)));
item_set_int(t,"line_number",ts->line_number);
}
delete_utterance(utt);
ts_close(ts);
return durs;
}
